The Chloride Problem at the Heart of Epilepsy

Mesial temporal lobe epilepsy (mTLE) is the most common focal epilepsy in adults, and it is stubbornly resistant to treatment. Surgical removal of the epileptic focus controls seizures in roughly 60 to 75 percent of patients at two years, but fewer than half remain seizure-free a decade later. For the millions who cannot be operated on, or for whom surgery fails, the therapeutic cupboard is nearly bare.

The biological root of this resistance runs deeper than most current drugs can reach. In healthy neurons, a protein called KCC2 (encoded by the gene SLC12A5) acts as a molecular pump, continuously expelling chloride ions from the cell. This keeps intracellular chloride low, which in turn ensures that the neurotransmitter GABA does its proper job: when GABA opens chloride channels, chloride rushes in, hyperpolarizing the neuron and suppressing its activity. In mTLE, KCC2 expression and function are frequently reduced. Chloride accumulates inside neurons. GABA, paradoxically, becomes excitatory. The very system meant to brake runaway electrical activity instead accelerates it.

Blocking the opposing chloride importer NKCC1 with the diuretic bumetanide was one proposed fix, but two clinical trials in neonatal seizures failed, and the drug's broad peripheral expression generates unwanted side effects. KCC2, by contrast, is expressed almost exclusively in neurons, making it a far cleaner therapeutic target. The question was whether anyone could find a drug that actually works on it.

65 million people worldwide live with epilepsy. mTLE, the most common adult focal form, is frequently refractory to all available medications.

From Rat Neurons to Human Tissue to Living Mice

The Paris-based team led by Jean Christophe Poncer at the Institut du Fer à Moulin built their case across four experimental layers, each one closer to the clinic than the last.

Key Method: Measuring KCC2 FunctionKCC2 activity was assessed in cultured rat hippocampal neurons by measuring the somato-dendritic gradient of the GABA reversal potential (ΔEGABA) using whole-cell patch-clamp recordings. Because KCC2 extrudes chloride more efficiently in dendrites than at the soma, a steeper gradient indicates stronger transporter function. A specific KCC2 antagonist nearly abolishes this gradient; bumetanide does not touch it. This makes ΔEGABA a reliable, specific readout of KCC2 activity.
Key Method: Tracking Single TransportersTo watch KCC2 move in real time, the team attached quantum dots to a recombinant Flag-tagged version of the transporter expressed in hippocampal neurons. Single-particle tracking revealed how freely the transporter diffuses within the plasma membrane and how often it settles into stable clusters, both near and away from inhibitory synapses.
Key Method: Human Tissue RecordingsAcute slices from surgically resected temporal lobe tissue from 13 mTLE patients were placed on multi-electrode arrays (MEAs). Spontaneous interictal-like discharges (IILDs), the rhythmic bursts of synchronized activity that characterize epileptic tissue, were recorded before and after drug treatment. Eleven of the 13 patients had hippocampal sclerosis, the canonical pathology of mTLE.
Key Method: In Vivo Mouse ModelStatus epilepticus was induced in mice with lithium and pilocarpine. After a latent period of three to four weeks, spontaneous seizures emerged. Telemetric electrocorticography (ECoG) recorded seizures, interictal discharges (IIDs), and high-frequency oscillations (HFOs, above 80 Hz) continuously. Mice received twice-daily injections of either PCPZ (2 mg/kg) or saline for five days, and seizure metrics were compared to a five-day baseline period.

Two Drugs, One Target, Confirmed

The first order of business was settling a controversy. Prochlorperazine (PCPZ), a first-generation antipsychotic approved for nausea and schizophrenia, and CLP-257, a compound from the arylmethylidine family, had each been proposed as KCC2 enhancers based on cell-line screening. But their efficacy in actual cortical neurons had never been rigorously established, and a 2017 paper in Nature Medicine argued that CLP-257 did not touch KCC2 at all, acting instead on GABA-A receptors.

The new data resolve this cleanly. After a two-hour incubation with PCPZ at 10 µM, the somato-dendritic ΔEGABA gradient in rat hippocampal neurons rose from 6.6 ± 0.8 mV/100 µm to 11.5 ± 1.5 mV/100 µm (p = 0.006). CLP-257 at 100 µM produced a comparable shift, from 5.3 ± 1.1 to 9.7 ± 1.1 mV/100 µm (p = 0.014). Both compounds are bona fide KCC2 enhancers in neurons.

Fig. 1. Both PCPZ and CLP-257 increase the somato-dendritic EGABA gradient in rat hippocampal neurons, confirming their action as KCC2 enhancers. Bar graphs show the ΔEGABA for control versus...
Fig. 1. Both PCPZ and CLP-257 increase the somato-dendritic EGABA gradient in rat hippocampal neurons, confirming their action as KCC2 enhancers. Bar graphs show the ΔEGABA for control versus drug-treated neurons; I-V curves illustrate the shift in reversal potential between soma and dendrite.
"Both PCPZ and CLP-257 effectively potentiate KCC2 function in rat hippocampal neurons."— Donneger et al., 2023

A Dual Action That Matters for Interpretation

The controversy about CLP-257 turns out to be a case where both sides were right. CLP-257 does enhance KCC2, and it also potentiates GABA-A receptors, specifically the extrasynaptic variety. When CLP-257 was applied acutely, the decay time constant of currents evoked by focal GABA agonist application increased from 124.4 ± 9.1 ms to 218.2 ± 21.1 ms (p < 0.001), a benzodiazepine-like prolongation of inhibitory current. This effect appeared at both 100 µM and 10 µM, suggesting it is not dose-dependent in the range tested.

PCPZ showed none of this. Its effect on GABA-A current amplitude and decay was statistically indistinguishable from vehicle. Neither drug altered miniature inhibitory postsynaptic currents (mIPSCs), the readout for synaptic GABA-A receptors. CLP-257 acts on extrasynaptic but not synaptic GABA-A receptors; PCPZ does not act on GABA-A receptors at all. This distinction matters enormously for interpreting any downstream effects on network activity, and it makes PCPZ the cleaner tool for attributing antiepileptic effects specifically to KCC2 enhancement.

Fig. 2. CLP-257 prolongs the decay of extrasynaptic GABA-A receptor currents (panels C-D), while PCPZ has no detectable effect on either evoked or spontaneous inhibitory currents (panels A-B,...
Fig. 2. CLP-257 prolongs the decay of extrasynaptic GABA-A receptor currents (panels C-D), while PCPZ has no detectable effect on either evoked or spontaneous inhibitory currents (panels A-B, E-H). Neither drug alters synaptic mIPSC kinetics.

Not More Protein. Better Organized Protein.

The most mechanistically surprising finding concerns how these drugs actually boost KCC2 function. The obvious hypothesis would be that they increase the amount of transporter on the cell surface. Surface biotinylation assays ruled this out. After two hours of treatment with either compound, total KCC2 expression was unchanged, and the ratio of surface to total KCC2 was unchanged. The drugs are not trafficking more protein to the membrane.

Single-particle tracking told a different story. KCC2 normally wanders laterally within the plasma membrane, diffusing in and out of functional clusters. Both PCPZ and CLP-257 curtailed this wandering. PCPZ reduced the extrasynaptic diffusion coefficient of KCC2 by 36.5 ± 14.9% (p < 0.001) and the explored membrane area by 39.7 ± 17.2% (p < 0.001). CLP-257 produced comparable reductions extrasynaptically. The transporter was not disappearing from the surface; it was being corralled.

Immunocytochemistry confirmed the consequence: PCPZ increased KCC2 cluster intensity by 109.0 ± 61.4% and cluster area by 45.3 ± 18.1% (both p < 0.001). CLP-257 produced smaller but still significant increases in both metrics. Cluster density, the number of clusters per unit area, did not change for either drug. The picture that emerges is a redistribution of existing membrane-inserted KCC2 into larger, denser functional aggregates, without any net change in how much transporter is present.

Fig. 3. PCPZ and CLP-257 reduce KCC2 lateral diffusion (panels C-D) and increase its clustering (panels E-F) in hippocampal neurons, without altering total or surface KCC2 expression levels...
Fig. 3. PCPZ and CLP-257 reduce KCC2 lateral diffusion (panels C-D) and increase its clustering (panels E-F) in hippocampal neurons, without altering total or surface KCC2 expression levels (panels A-B). Single-particle tracking trajectories show more confined movement after drug treatment.
+109% increase in KCC2 cluster intensity after PCPZ treatment, with no change in total or surface KCC2 protein levels. The transporter is reorganized, not upregulated.

The Known Switches Were Not Flipped

KCC2 function is regulated by phosphorylation of several key residues in its carboxy-terminal domain. Phosphorylation of serine 940 (S940) by protein kinase C stabilizes the transporter and boosts its activity. Phosphorylation of threonines 906 and 1007 (T906, T1007) by the WNK/SPAK/OSR1 kinase cascade does the opposite, destabilizing KCC2 and reducing chloride extrusion. These are the canonical levers for KCC2 modulation, and they were the obvious candidates for how PCPZ and CLP-257 might work.

Immunoprecipitation assays on adult rat hippocampo-cortical slices showed that neither drug altered phosphorylation at S940, T906, or T1007. The WNK inhibitor WNK463, used as a positive control, reduced pT1007 by 78.7 ± 3.3% and pSPAK by 84.1 ± 4.6%, confirming the assay was working. PCPZ and CLP-257 moved none of these needles. Tyrosine phosphorylation of KCC2, another regulatory mechanism, was similarly unaffected. In vitro kinase assays confirmed the drugs do not directly inhibit WNK1, WNK3, SPAK, or OSR1.

The mechanism of KCC2 enhancement by these compounds is therefore independent of all canonical phosphorylation pathways. What they are doing at the molecular level remains an open question. The authors suggest several possibilities: interaction with alternative phosphorylation sites identified in proteomics studies, interference with KCC2's protein-protein interactions (candidates include gephyrin, Neto2, and SNAP23), or promotion of KCC2 accumulation into lipid rafts. None of these has been tested directly.

Fig. 4. Phosphorylation of canonical KCC2 regulatory residues (S940, T906, T1007) is unaffected by PCPZ or CLP-257 treatment, as shown by immunoprecipitation and western blotting from adult rat...
Fig. 4. Phosphorylation of canonical KCC2 regulatory residues (S940, T906, T1007) is unaffected by PCPZ or CLP-257 treatment, as shown by immunoprecipitation and western blotting from adult rat hippocampo-cortical slices. The WNK inhibitor WNK463 serves as a positive control, dramatically reducing pT1007 and pSPAK levels.

Silencing Storms in Human Epileptic Tissue

The translational centerpiece of the study is the human tissue data. Brain slices from 13 patients who had undergone surgical resection for mTLE were placed on multi-electrode arrays and allowed to generate their characteristic spontaneous interictal-like discharges. These IILDs, low-frequency synchronized bursts below 40 Hz, are the electrophysiological signature of epileptic tissue and a validated surrogate for seizure propensity in this preparation.

In slices from five patients with typical ILAE type 1 hippocampal sclerosis, PCPZ virtually abolished IILDs, producing a 97.6 ± 13.8% reduction in discharge frequency (p < 0.001). CLP-257 was equally effective in five of six patients tested, reducing IILD frequency by 96.5 ± 2.3% (p < 0.001). Crucially, this suppression did not reflect neuronal death or silencing: when excitability was boosted with high-potassium, low-magnesium solution, both IILDs and multi-unit activity returned, confirming that neurons remained viable and capable of firing.

Across all 13 patients, both drugs significantly reduced the proportion of slices displaying spontaneous IILDs (p < 0.05 for both). The effect was at least comparable to, and in many cases stronger than, the partial suppression previously reported for bumetanide in similar preparations.

Fig. 5. Multi-electrode array recordings from human mTLE postoperative tissue. Spectrograms show the near-complete suppression of spontaneous interictal-like discharges (IILDs) after PCPZ...
Fig. 5. Multi-electrode array recordings from human mTLE postoperative tissue. Spectrograms show the near-complete suppression of spontaneous interictal-like discharges (IILDs) after PCPZ treatment compared to DMSO control and partial suppression by bumetanide. The bar graph (panel D) shows the proportion of slices with IILDs across conditions; panel G shows patient-level responses, with non-responders highlighted.
"Enhancing KCC2 function prevents epileptiform activity recorded in vitro in postoperative tissue from a majority of patients with mTLE and typical ILAE type 1 HS."— Donneger et al., 2023

When the Drug Does Not Work: A Diagnostic Signal

Three tissue samples did not respond. One patient had ILAE type 3 hippocampal sclerosis, a rarer pattern often associated with dual pathology. Another had a ganglioglioma carrying a somatic BRAF V600E mutation, a tumor-driven epilepsy with a fundamentally different biology from sclerosis-associated mTLE. A third showed partial non-response in one slice.

This pattern of non-response is not a failure of the approach; it is a diagnostic signal. The data suggest that KCC2 enhancement is specifically effective when the underlying pathology involves the chloride transport dysfunction characteristic of hippocampal sclerosis. When the epilepsy arises from a different mechanism, restoring KCC2 function may simply be irrelevant. Whether the non-responding tissues have preserved KCC2 expression, or whether their epileptiform activity is driven by mechanisms entirely upstream of chloride homeostasis, remains to be determined.

Forty Percent Fewer Seizures in Living Animals

Human tissue recordings are powerful but cannot capture the full complexity of spontaneous, recurrent seizures. For that, the team turned to the lithium-pilocarpine mouse model, which recapitulates the key features of mTLE: status epilepticus followed by a latent period, then chronic spontaneous seizures, hippocampal sclerosis, granule cell dispersion, mossy fiber sprouting, and reduced KCC2 immunoreactivity in the dentate gyrus.

After a five-day baseline recording period, mice received twice-daily injections of PCPZ at 2 mg/kg or saline for five days. Seizure frequency in saline-treated animals was unchanged between baseline and test periods (0.15 ± 0.07 vs. 0.18 ± 0.06 seizures per hour, p = 0.118). In PCPZ-treated animals, seizure frequency fell from 0.14 ± 0.05 to 0.09 ± 0.05 seizures per hour, a reduction of approximately 40% (p = 0.01). Seizure duration and power were unaffected, pointing to an effect on seizure initiation rather than propagation or termination.

The effects on interictal biomarkers were even more pronounced. Interictal discharge frequency fell by 34% with PCPZ (from 290.9 ± 60.3 to 192.0 ± 39.3 IID per hour, p = 0.01), while high-frequency oscillation frequency dropped by 58% (from 128.68 ± 71.88 to 54.61 ± 23.73 HFO per hour, p = 0.012). In saline controls, HFO frequency actually increased by 38% over the same period, suggesting a natural progression of epileptic network activity that PCPZ arrested. The effect was consistent across both light and dark phases of the circadian cycle, ruling out a simple interaction with sleep-wake rhythms.

Fig. 6. Chronic PCPZ administration in the lithium-pilocarpine mTLE mouse model. Telemetric ECoG recordings show reduced seizure frequency (panel F), interictal discharge frequency (panel G), and...
Fig. 6. Chronic PCPZ administration in the lithium-pilocarpine mTLE mouse model. Telemetric ECoG recordings show reduced seizure frequency (panel F), interictal discharge frequency (panel G), and high-frequency oscillation frequency (panel H) during the PCPZ treatment period compared to the saline control period. Histology confirms hippocampal sclerosis and reduced KCC2 immunoreactivity in the dentate gyrus of epileptic mice (panels C-D).
58% reduction in high-frequency oscillations (HFOs) with chronic PCPZ. HFOs are a validated biomarker of epileptogenic networks and a target for next-generation antiepileptic strategies.

What the Study Establishes and What It Leaves Open

The breadth of this work is its greatest strength. Moving from single-transporter tracking in cultured neurons to multi-electrode recordings in human surgical tissue to chronic telemetric monitoring in living animals, within a single paper, is genuinely unusual. Each layer of evidence reinforces the others, and the convergence on a consistent mechanism, membrane confinement and clustering of KCC2, is compelling.

The mechanistic gap is real, though. Showing that canonical phosphorylation pathways are not involved narrows the field but does not identify the actual target. The drugs could be acting on any of several protein-protein interactions that regulate KCC2 clustering, or on lipid raft dynamics, or on phosphorylation sites not yet characterized. Until the binding target is known, rational optimization of these compounds is constrained.

The dual action of CLP-257 also deserves scrutiny. Its potentiation of extrasynaptic GABA-A receptors is a benzodiazepine-like effect, and in a tissue where GABA is paradoxically excitatory due to chloride accumulation, potentiating GABA-A receptors could theoretically worsen activity rather than suppress it. The fact that CLP-257 still suppresses IILDs in human tissue suggests the KCC2-enhancing effect dominates, but the two actions cannot be cleanly separated in any functional assay. PCPZ, which lacks this confound, is the more interpretable tool.

The in vivo PCPZ data carry a caveat the authors address directly: PCPZ is a dopamine D2 receptor antagonist, and one could ask whether D2 antagonism rather than KCC2 enhancement drives the antiseizure effect. The authors argue against this on pharmacological grounds, noting that D2 agonists, not antagonists, display antiepileptic properties in the literature, and that the D2 antagonist eticlopride did not affect KCC2 clustering in their assays. The argument is reasonable but not definitive; a genetic or pharmacological rescue experiment using a KCC2-specific intervention would close this loop.

A New Blueprint for Drug-Resistant Epilepsy

The clinical implications are worth stating plainly. PCPZ is already FDA-approved, has a known safety profile in adults, and crosses the blood-brain barrier. The dose used in mice (2 mg/kg) is within a range that primarily engages D2 receptors without motor side effects in rodents. Whether the antiepileptic effect in humans would require doses that produce extrapyramidal symptoms or other adverse effects is unknown, and that question will need to be answered before any clinical translation. The authors note that adverse effects in children are a concern with PCPZ, which is relevant given that mTLE can present in adolescence.

The broader significance is conceptual. For decades, antiepileptic drug development has focused on ion channels, synaptic receptors, and neurotransmitter metabolism. This study adds a new category: chloride transporter stabilization. By showing that corralling KCC2 into functional membrane clusters, without changing how much of the protein is present, is sufficient to suppress seizures in human tissue and reduce them in living animals, the work establishes a proof-of-concept that will be difficult to ignore.

The recent resolution of KCC2's three-dimensional structure by cryo-electron microscopy opens the door to structure-based drug design. Compounds more specific than PCPZ or CLP-257, designed to bind defined sites on the transporter and promote its clustering without off-target receptor interactions, are now a realistic goal. The path from this preprint to a new class of antiepileptic drugs is long, but the first steps have been mapped with unusual clarity.